JP2002152049A - Data processing apparatus and data processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve modeling processing to enhance prediction accuracy by utilizing features of a data source. SOLUTION: The data processing apparatus of this invention is provided with a first prediction processing circuit where a data symbol of a data source having first component data and second component data relating to the first component data is expressed as a modeled data symbol, a data symbol of the first component is predicted from preceding first and second component data symbols to generate the modeled data symbol to express the first component data symbol. A difference between the preceding first component data symbol and the preceding second component data symbol corresponding to the preceding first component data symbol is calculated and the difference is subtracted from the second component symbol corresponding to the first component symbol to decide the predicted first component symbol, the predicted value corresponding to the first component symbol is subtracted from the first component symbol to calculate a prediction error of the first component symbol and the modeled data symbol is generated from the predicted error.

Description

DETAILED DESCRIPTION OF THE INVENTION

Description: FIELD OF THE INVENTION The present invention relates to transforming data symbols from a data source into a representation of the information content of the source data and having a different probability of occurrence than the original source data symbol. Transformed data symbol
To a device and method for converting to

[0002] Further, the present invention relates to a data compression encoder and a data compression decoder.

BACKGROUND OF THE INVENTION It is an object of the present invention to convert a source data symbol into another type of converted data symbol having a different probability of occurrence from the original source data symbol and representing the same information content as the original source data. There are various applications to do. The process of converting to a data symbol having a different occurrence probability from the data symbol from the source is performed by modeling (mo
This is called delling or pre-processing. Modeled data symbol
Is typically less redundant than the original source data symbol.

An example of a modeling application is in the field of data compression coding. The data compression encoder compresses a fixed amount of source data into compressed encoded data having a substantially reduced data amount. As the compression encoding method, there is a method in which no information is lost when the source data is compressed and encoded, and a method in which information is intentionally reduced to increase the compression encoding efficiency. Examples of coding processes without loss of information include the Joint Photographic Experts Group:
Hereinafter, it is called JPEG. 2.) There is an encoding process, which is usually applied to the digital representation of a still image captured by, for example, a digital video camera. In the JPEG encoding process, data compression is effectively performed using Huffman encoding. Huffman coding is one of the compression coding algorithms that is effectively performed by preprocessing or modeling the source data. Compared to other known compression coding algorithms, Huffman coding provides the most efficient compression of source data with lower entropy. The purpose of the modeling or preprocessing is to convert the data source symbols into modeled data with lower entropy.

In the lossless JPEG example, the modeling process involves differential pulse code modulation (Differen
tial Pulse Code Modulation: Hereinafter, referred to as DPCM. ) Also known as modeling. In DPCM modeling, a model that generates an estimate of each of a plurality of preceding source data symbols weighted by corresponding weighting factors and represents a prediction error based on the difference between each predicted data symbol and the original data symbol. By generating a new data stream having encoded data symbols, the entropy of the source data is reduced.

Usually, the closer the prediction of each source data symbol generated by the preprocessing is to the original source data symbol, the more effectively the modeling preprocessing is executed.
Therefore, it is desired to realize a modeling process for generating a modeled data symbol for a data source, in which the characteristics of the data source are used to improve the accuracy of prediction.

DISCLOSURE OF THE INVENTION In order to achieve the above object, a data processing apparatus according to the present invention uses a data symbol of a data source having data of a first component and data of a second component related thereto as a model. In a data processing device represented as a modeled data symbol, a first data symbol of each first component is predicted from a data symbol of a preceding first component and a data symbol of a second component. And a first prediction processing circuit that generates a modeled data symbol representing the data symbol of the component.

The data processing apparatus according to the present invention utilizes a predetermined degree of correlation existing between occurrence probabilities of data symbols from each of a plurality of components of a data source. By predicting one of the components from the other components, an improved modeling process can be performed. For example, the components of a color image are composed of red, green, and blue pixels.
In a normal color image, red pixels, green pixels, and blue pixels have a correlation. That is, it can be said that there is a certain degree of relationship between the pixel values of each component. Accordingly, embodiments of the present invention utilize this correlation, for example, by predicting a red component from a green component.

In a preferred embodiment of the present invention, the first prediction processing circuit calculates a difference between a preceding symbol of the first component and a corresponding symbol of the preceding second component, By subtracting this difference from the second component symbol corresponding to each first component symbol, the predicted value of each first component symbol is determined, and the first component symbol is converted to the first component symbol. By subtracting the predicted value corresponding to the symbol of the first component, a prediction error of each first component symbol is calculated, and a modeled data symbol is generated from the prediction error.

Preferably, in order to reduce the alphabet size of the modeled data symbol, the prediction processing circuit generates a modeled data symbol from each prediction error modulo the alphabet size of the source data symbol. The prediction error is calculated by subtracting the current symbol value from the predicted value. As a result, if the current symbol has N possible values, the prediction error is 2N
-1 will have one possible value. However,
As described above, by using the alphabet size of the source data symbol as a modulus, the number of possible values of the prediction error can be reduced to this alphabet size of N.

A data processing apparatus according to the present invention generates a first modeled data symbol by predicting a first data symbol from a second data symbol, but generates a modeled data symbol representing a second component. It is often desirable to generate a data symbol, which allows all information from the data source to be transformed into a modeled data symbol having a different probability of occurrence than the probability of occurrence in the source data symbol. For this reason, the data processing device according to the present invention predicts each second component data symbol from the preceding second component data symbol, and calculates the second symbol based on the difference between the original second data symbol and this prediction. May be provided with a second prediction processing circuit that generates the modeled data symbol.

In a preferred embodiment of the present invention,
The second prediction processing circuit performs differential pulse code modulation (Differ
It operates on the basis of the Essential Pulse Code Modulation (DPCM).

Further, the present invention provides a data compression encoding apparatus for preprocessing source data having two or more components and compressing and encoding modeled data symbols by a compression encoding process. Further, the present invention provides a data compression / decoding device corresponding to the data compression / encoding device.

In the embodiment of the present invention, Huffman coding is used as a specific example of the data compression coding algorithm, but the present invention is not limited to a specific data compression coding system. That is, the present invention can be applied to both the data compression encoding method without loss and the compression encoding method in which part of information is lost.

Further, the present invention relates to a method for detecting color images in a color image.
An improved data modeling pre-process for predicting one component from two other components is provided.

[0016] Further aspects and features of the present invention are defined in the appended claims.

DESCRIPTION OF THE PREFERRED EMBODIMENTS As an embodiment of the present invention, a data source (data source) is used.
modeling process (mode) applicable to applications where it is beneficial to change the probability of occurrence of data symbols from e)
lling process). An example of such an application field is data compression encoding.
By pre-processing the data symbols of the source data by the modeling process, the modeled data symbols (mode
The probability of occurrence of an lled data symbol) allows the compression encoding process to be executed more efficiently.

FIG. 1 is a diagram comprehensively showing a configuration in which source data is compression-encoded, compression-decoded, and supplied to a data sink. As shown in FIG. 1, a data source 1 supplies a source data symbol to a data compression encoder 2. This source data symbol is input to the pre-processing circuit 4 of the data compression encoder 2 via the connection channel 6. The data compression encoder 2 includes a compression encoding processing circuit 8, and the preprocessing circuit 4 supplies the preprocessed data symbols to the compression encoding processing circuit 8. The compression encoding processing circuit 8 generates compression encoded data (compression encode data) corresponding to the source data.
d data) to the output channel 10. The compression encoding algorithm executed by the compression encoding processing circuit 8 compresses the source data to generate compressed encoded data whose data amount is greatly reduced. The compression encoded data is
It is supplied to a channel or recording medium, shown generically as box 12. In order to explain the data compression encoding process, FIG. 1 also shows a process of restoring the compression encoded data. The compressed encoded data is supplied from the channel or the recording medium 12 to the data compression decoder 14 via the connection channel 16. The data compression decoder 14
A compression / decoding processing circuit 18 and a post-processing circuit 20 connected to the compression / decoding processing circuit 18 are provided. Post-processing circuit 20
Receives the compressed and decoded data symbols from the compression and decoding processing circuit 18 via the connection channel 22. The post-processing circuit 20 performs a process reverse to that of the pre-processing circuit 4 of the data compression encoder 2, and supplies an estimated value (source) of the source data to the sink 26 via the output channel 24.

As a preferred embodiment of the present invention, a specific example in which the present invention is applied to digital image encoding will be described. However, the present invention can be applied to other types of data.

FIG. 2 shows a detailed configuration of the data compression encoder 2 shown in FIG. In FIG. 2, the same parts as those in FIG. 1 are denoted by the same reference numerals. In a specific example where the source data represents a color digital image,
The data compression encoder 2 needs to effectively encode three different source data streams representing the red, green and blue components of the color image. Therefore, the data compression encoder 2 shown in FIG. 2 performs preprocessing and compression encoding on the three components of the color image. Therefore, the preprocessing circuit 4 shown in FIG. 2 includes three data processors 30, 32, and 34 that receive data symbols corresponding to three components of red, green, and blue of a color image, respectively. Correspondingly, the compression / encoding processing circuit 8 also includes three processors 36, 3
8, 40, and these processors 36, 38, 40
Include the data processors 30, 32, 3 of the preprocessing circuit 4.
The data symbols pre-processed by 4 are supplied via output channels 33, 35 and 37, respectively.

The data processor 30 has a modeler (modell).
er), and has the function and purpose of converting the source data symbols supplied via the connection channel 6 into a new stream of modeled data symbols with low entropy. As is well known to those skilled in the art, the term entropy as used for an information source means a measure of the relative amount of information provided by the symbol of that source. Assuming that i = 1 to n and the symbol occurrence probability of the data source 1 is p _i , the entropy of the data source is calculated by the following equation (1).

[0022]

(Equation 1) As a result of the preprocessing, the data source is modeled, and the entropy of the data source is reduced. As a result, more efficient data encoding can be performed in the data compression encoding process following the preprocessing. The reason is that the data compression encoding algorithm compresses data sources that have symbols that occur in a "peaky" biased probability distribution, rather than a flat distribution in which each data symbol occurs with equal probability. This is because the ratio increases. This will be described in more detail in a later section.

In this example, the data source is:
For generating a digital image, the preprocessing circuit 4 converts the symbol of each component of the color image into a modeled data symbol. As described later, the three components of the color image are modeled by different methods. This is because, in this specific example, a correlation that always exists between the three color components of the color image is used. This process will be described with reference to FIGS. 3 and 4 showing two test images by graphic representation of the pixel values of the three components of the color image. FIGS. 3 and 4 show graphs in which the pixel positions and the pixel values are plotted for R, G, and B, which are the red, green, and blue components of each of the two test images. As can be seen from the two test images shown in this example, three components R,
The values of G and B are close to each other. For example, in the specific example shown in FIG. 4, peaks 41 and 4 of the pixel values of each of the three components.
2, 43 and 44 are present at close positions.
That is, it is understood that the pixel values of the image have a strong correlation. Utilizing this correlation in the modeling preprocessing, by predicting one of the three components from the other two, a modeling that has lower entropy and therefore can be more efficiently compression encoded Data symbols can be generated. Hereinafter, such modeling processing is referred to as component differential prediction (Component Differential P
rediction: Hereinafter, referred to as CDP. ) Called modeling. One component serving as a reference for prediction is called a reference component, and in this specific example, a red component is used as a reference component. It is clear that other components may be used as reference components.

FIG. 5 shows details of the preprocessing circuit 4 for performing modeling in the data compression encoder 2 shown in FIG. In addition,
In FIG. 5, the same reference numerals are given to the same circuits as those shown in FIGS. 1 and 2. FIG. 5 shows the data processors 30, 32, and 34 in more detail.
In this example, the data processor 34 is a single processor, and the other two data processors 3
0, 32 are three more processing units 95, 97, 9
9. In this embodiment, the reference component red component R is supplied to the data processor 34, and the other components green component G and blue component B are supplied to the other two data processors 30, 32, respectively. Is done.
In order to explain the operation of the preprocessing circuit 4, first, a modeling technique applied to a red component R as a reference component will be described. The reference component needs to be able to be supplied from the modeled data symbol representing the data symbol of the reference component to the data compression decoder 14 without reference to the other two components. Therefore, the data processor 34 assigned to the red component executes an operation based on the DPCM modeling process described later with reference to FIGS. A more detailed description of the DPCM modeling process and a variable weight DPCM modeling process (Variable Weigh
The tDPCM modeling process is also disclosed in co-pending UK Patent Application No. 0014890.8, which is incorporated herein by reference.

As shown in FIG. 6, the data processor 34
Processes pixels corresponding to red representing the red component of the digital image 46. In FIG. 6, the lines 46 and 50 indicate the image 46.
Is enlarged and shown as a group of pixels in box 54. The box 54 is made up of a plurality of squares 56, each of which represents a pixel corresponding to red in some of the images 48. Box 54
A line 58 shown inside represents a part of the object in the image 46. In this example, line 58 is a tree 60
Part of. As can be seen from the magnified view of image 48 shown in FIG. 6, most of the pixels in some images 48, i.e., the pixels other than the pixels forming line 58, have the same relative amplitude and therefore have the same Have. This is a result of the normal image having a large area corresponding to the same pixel value, and the data processor 34 performing the preprocessing uses this feature to reduce the entropy of the source data.

As shown in FIG. 7, the data processor 34 performing the preprocessing advances the processing for each column in order from the upper left pixel 60 to the lower right pixel 62. Here, it is assumed that the preprocessing data processor 34 has already processed the pixel at the position 64, is going to process the value of the pixel at the position 66, and has not yet processed the pixel at the position 68. Since x, y, and z are known, a predicted value pa of the value of _a can be obtained from them by, for example, the following equation.

[0027]

(Equation 2) Or

[0028]

(Equation 3) Alternatively, a comprehensive linear prediction process (general linear pre
dictor) _is, as _{w x + w y + w z} ≠ 0, the weights _w x,
It can also be performed based on the following equation using w _y and w _z .

[0029]

(Equation 4) Of course, it can also be used to predict other pixels.
In fact, a simple one-dimensional prediction process (one-dimension predictor) can be performed based only on the immediately preceding pixel. However, two-dimensional prediction processing usually requires 1
Better than dimensional prediction processing. The prediction process shown in Expression (4) is the simplest form of the two-dimensional prediction process.

Here, some considerations must be taken.
There are special cases. -For the pixel at the upper left corner of the image, this pixel is processed
Cannot be predicted because it is the first pixel
No. In this case, the predicted value is set to 0 (here, other
Pixel values may be used). -Only one-dimensional prediction processing is possible for the pixels in the top row
It is. For example, p_a= X etc. are sufficient. -For the pixel on the leftmost row, the pixel on the left
And therefore always p _a= Y etc. are predicted.

[0031] Once the predicted value is formed, it can be based on equation (5) to calculate the prediction error e _a.

[0032]

(Equation 5) The prediction error e _a is used to generate a modeled data symbol, is output from the data processor 34 for performing modeling, and is output from the processor 4 for performing compression encoding.
0 is supplied. If the data processor 34 for preprocessing has performed a good pretreatment values close e _a frequency count number 0 (frequency counts) becomes very large. This pre-processing is useful for coding, since data streams with only a small number of frequent symbols, as opposed to all symbols with similar probabilities, are well compressed. However, this process increases permissible symbols. For example, N
Assuming that there are a number of possible pixel values (having a value in the range of 0 to N), the DCPM modeling process increases the alphabet size, resulting in 2N-1 possible values. I will. This phenomenon, for a given predicted value p _a, the possible values for the prediction error can be avoided by the N pieces. That is, as shown in Expression (6), the prediction error can take a value modulo N.

[0033]

(Equation 6) This means that it is not necessary to increase the alphabet size by the DCPM modeling process. As described later, the post-processing circuit 20 performs an inverse modeling process. Conversely modeling process, based on the predicted value p _a of the pixel values a, is realized by executing the processing of the preprocessing circuit 4 and the reverse. That is, the post-processing circuit 20 calculates the value v = (a
-P _a) receiving a mod (N), performs the inverse modeling process. The pixel value is calculated by equation (7).

[0034]

(Equation 7) Next, the processing for the green component and the blue component by the remaining two data processors 30 and 32 will be described.

Based on the correlation between the three components of the color image, the data processor 30, 32 generates a stream of modeled data symbols for the blue and green component input data streams, respectively, with respect to the red component input data stream. Generate. Data processor 30, 3
2 is a correlation evaluator
95. The correlation evaluator 95 is supplied with the data symbol of the red component via the connection line 70 and the data symbol of the green component and the blue component for which the modeled data symbol is to be calculated. Hereinafter, processing for the green component will be described, but it is apparent that the blue component can be processed in the same manner. Therefore, a duplicate description will not be given. The correlation evaluator 95 calculates a correlation value for each green pixel from the difference between the value of the preceding green pixel and the value of the preceding red pixel. This process can be expressed as equation (8).

[0036]

(Equation 8) Subsequently, the correlation value (diff) of each green pixel is calculated by a predictor 97.
Supplied to The predictor 97 calculates a predicted value of each green pixel by subtracting the correlation value (diff) from the corresponding red pixel value based on the correlation value of the green pixel and the corresponding red pixel value. This process can be expressed as equation (9).

[0037]

(Equation 9) Next, the error in the predicted value is provided to the encoder. This error is generated in the error predictor 99. The error predictor 99 receives the predicted value of each green pixel via a connection channel 76 and receives the original green pixel value via another connection channel 78. The error predictor 99 calculates a prediction error for each green pixel based on Expression (10).

[0038]

(Equation 10) The error is converted to a value modulo the alphabet size of the pixel value N. Like the DPCM modeling process, this process is performed to avoid increasing the alphabet size to 2N-1. Note that the values R ＾, G ＾, and R are supplied to the modeling process and the inverse modeling process performed by the post-processing circuit 20 in the decoder, and the pre-processing circuit 4 and the post-processing circuit 20 determine the value of G _pred. Therefore, the processing using the pixel value N as a modulus can be performed here. Therefore, the value that GG _{pre d} can take is N
There are only pieces.

The data compression decoder 14 performs an inverse modeling process, which is an inverse process of the modeling process, by performing a compression decoding process and a post-process on the modeled data symbol restored by the compression decoding process. The inverse modeling process for the green component is performed using the formula (1) for each pixel.
This is performed by applying 1).

[0040]

[Equation 11] The data compression / decoding processing circuit 18 shown in FIG. 1 performs the reverse processing of the data compression / encoding processing, and the post-processing circuit 20 performs the inverse modeling processing. For the red component, this process is DP
This is the reverse process of the CM modeling process. For the green and blue components, this process is the reverse of the CDP modeling process. The data compression decoder 14 is shown in more detail in FIG. In FIG. 8, the same reference numerals are given to the same circuits as those in FIG.

Like the data compression encoder 2, the data compression decoder 14 processes a color image signal by dividing it into red, green and blue components. The encoded portions of the image signal are supplied to decoding processors 130, 132, and 134 that perform the compression and decoding that constitute the compression and decoding processing circuit 18, respectively. The decoding processors 130, 132, and 134 execute, for example, the reverse process of the compression encoding algorithm, and estimate the modeled data symbol (estimate) for the red, green, and blue components input to the data compression encoder 2. ).

As shown in FIG. 8, the modeled data symbols for the red, green and blue components are supplied to the post-processing units 142, 144 and 146 via the corresponding connection channels 136, 138 and 140, respectively. . The post-processing unit 146 performs an inverse DPCM modeling process based on Expression (7). The green component and the blue component are supplied to other post-processing units 142 and 144. These are respectively a correlation evaluator 151, a prediction processing circuit 153,
An output circuit 155 is provided. Like the pre-processing circuit, the post-processing units 142 and 144 for the green component and the blue component are the same, and the post-processing unit 1 for the green component is again used.
Only the process at 44 will be described.

The details of the post-processing unit 144 are shown in FIG. Here, in FIG. 9, the same circuits as those in FIGS. 1 and 8 are denoted by the same reference numerals. In order to perform the inverse process of the CDP modeling process, the modeling symbol of the image component is supplied to the correlation evaluator 151. Post-processing circuit 1
The estimated value of the red pixel generated by 46 is required for the process of reproducing the green pixel, and the estimated value of the red pixel is supplied to the correlation evaluator 151 via the connection channel 157. The correlation evaluator 151 calculates, based on the above equation (8),
Generate a correlation value for each modeled data symbol.
This correlation value is generated using the estimated pixel values provided via feedback channels 159, 159 'from output circuits 155, 155'. Here, the first estimated value of the green component G # is required. This estimated value is generated using an initial symbol value known to the post-processing circuit 20. The initial symbol value is known in the pre-processing circuit 4 and the post-processing circuit 20, and can be used to generate a first correlation value. The correlation value for each modeled data symbol is supplied to the prediction processing circuit 153. The prediction processing circuit 153 is also supplied with the estimated value of the pixel value of the red component from the post-processing circuit 146 via the connection channel 157. The prediction processing circuit 153 generates a predicted value of each green pixel value from each red pixel value by subtracting the correlation value from the red pixel value based on Expression (9).
This corresponds to an inverse modeling process for the modeled data symbol. Subsequently, the output circuit 155 calculates the expression (1)
Based on 1), an estimated value of each green pixel value is generated by adding the modeled data symbol corresponding to the predicted value of each green pixel value. Then, the estimated values of the data symbols of the red, green, and blue components are output from the connection channel 24.

Second Embodiment FIG. 10 shows a preprocessing circuit 4 according to a second embodiment of the present invention.
3 is a block diagram showing a modified configuration of FIG. Note that FIG.
, The same parts as those in FIG. 5 are denoted by the same reference numerals. The pre-processing circuit 4 shown in FIG. 10 performs further component difference modeling (Compon in which one image component among the three image components is calculated based on the other two image components).
It operates based on ent Differential Modeling (CDM) processing. The preprocessing circuit 4 in the second specific example includes the first processing circuit.
In order to perform an operation similar to the preprocessing circuit 4 described in the specific example, only the differences between these two specific examples will be described below. The first and second data processors 232 and 234 shown in FIG. 10 model the red component and the green component based on the DPCM processing, respectively. Therefore, in the post-processing circuit, the blue component can be calculated using both the red component and the green component. The blue component data processor 30 models the blue pixel values based on the assumption that the ratio of component differences between adjacent pixels will be the same. This estimation is represented by equation (12).

[0045]

(Equation 12) With this estimation, the predicted value of the blue pixel value can be generated based on the following pseudo code.

[0046]

(Equation 13)

[0047]

[Equation 14] The data processor 30 includes a prediction processing circuit 160 and an evaluation circuit 162. To execute this modeling process, the data processor 30 executes the process shown in FIG. 11 and described below. The process starts at step 200, and at step 202, the red, green, and blue pixels are supplied to the prediction processing circuit 160. For each blue data symbol (step 204), a first relation metric (G ＾ -B ＾) (step 206) and a first relation metric (R ＾ -G ＾) (step 208) are determined. A first relation metric is generated at step 206 from the difference between the preceding green pixel value and the preceding blue pixel value. A second metric is generated at step 208 from the difference between the preceding red and blue pixel values. Next, at step 210, for each blue pixel value, a third relation metric (RG) is generated as the difference between the corresponding red pixel value and the corresponding green pixel value. For each blue pixel value, at step 214
A determination is made whether the preceding red pixel value and the preceding green pixel value are equal. In the determination in step 214, the preceding red pixel value and green pixel value are equal (R ＾ =
G ＾), the predicted value of the blue pixel is the first relation metric (G ＾ -B) corresponding to the corresponding green pixel value.
＾) is calculated based on the difference. This processing is executed in step 216 by using the above equation (14). On the other hand, if the preceding red and green pixel values are not equal (R {G}), the predicted value of the blue pixel is the first green value selected by the corresponding green pixel value and the third relation metric.
And the ratio of the second relation metric. This processing is executed in step 218.

As shown in FIG. 10, the prediction value of each blue pixel is supplied to the evaluation circuit 162 via the connection channel 163. The evaluation circuit 162 is also supplied with a blue pixel value via the connection channel 165. Once the predicted values for each blue component have been generated, as described above, based on the prediction error formed as the difference between the blue pixel value and the corresponding blue pixel predicted value, modulo the pixel size alphabet size N, A modeled data symbol for the blue component is generated. This processing is performed in step 220.
This is executed based on the following equation (15).

[0049]

(Equation 15) The modeled data symbol is generated from the prediction error of each blue pixel by the evaluation circuit 162 based on Expression (15), and is supplied to the data compression encoding processing circuit.

FIG. 12 shows a post-processing circuit for performing the inverse modeling process in the second specific example. FIG.
In FIG. 2, the same circuits as those in FIG. 18 are denoted by the same reference numerals. The post-processing circuit calculates a blue component from the green component and the red component. Therefore, the two post-processing circuits 244, 246 have connection channels 138 ', 14
Via 0, modeled data symbols corresponding to the red and green pixel values, respectively, are provided. These two post-processing circuits 244, 246 perform the inverse DPCM modeling process described above to generate red and green pixel estimates. Subsequently, the post-processing circuits 244 and 246 output the estimated values of the red and green pixels. Note that the estimated values of the red pixel and the green pixel are also supplied to two input channels of the third post-processing circuit 242. Third post-processing circuit 242
Is also supplied with a modeled data symbol corresponding to the blue component, and the third post-processing circuit 242 performs an inverse CDP modeling process. The third post-processing circuit 242 includes a prediction processing circuit 260. The prediction processing circuit 260 calculates the equation (1)
A predicted value of each blue pixel value is generated based on 3), (14) and steps 200 to 218 of the flowchart shown in FIG. Here, instead of the original red and green pixels known in the encoder, the estimated values of the red and green pixels are used. The estimated value of each modeled data symbol is supplied to an output circuit 262. The output circuit 262 is also supplied with the modeled data symbols via the connection channel 136 '. Then, based on equation (15), each modeled data symbol corresponding to the blue pixel value is
An estimate of the blue pixel value is generated by adding the predicted value of the blue pixel. The estimated value of the blue pixel is output from the output channel 24.

As an embodiment of the present invention, an application for compressing and encoding an image has been described. However, the present invention is an application for performing compression in the form of data having a plurality of components. It can be applied to any application where a relationship exists. Various modifications can be made to the specific examples described herein without departing from the scope of the invention. In particular, signals representing data encoded according to the present invention are understood to be an aspect of the present invention.

As is apparent from the above description, the present invention also provides an image processing apparatus as an embodiment of the present invention.
The image processing apparatus compresses and encodes a source image including three components, that is, first, second, and third data, receives the first, second, and third components, and receives a third signal. A preprocessing circuit for generating first and second modeled data symbols representing first and second image components from the data of the image components; and a first and second modeling device connected to the preprocessing circuit. A compression encoding processor that compression-encodes the data symbol and the data symbol of the third component to generate a compression-encoded data symbol. The pre-processing circuit generates predicted values of the first component symbol and the second component symbol from the third component symbol, and calculates the predicted value of each first component symbol and the corresponding first component symbol. A first modeled data symbol is generated from the error between the predicted values and a symbol for each second component and a corresponding second
The second modeled data symbol is generated from the error between the predicted values of the symbol of the component. Here, the image is, for example, a color image, and the data of the first, second, and third components is
Represents the red, green, and blue components, respectively.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a general configuration of a data compression encoder and a data compression decoder.

FIG. 2 is a block diagram showing a configuration of a data compression encoder shown in FIG.

FIG. 3 is a diagram illustrating pixel values of red, green, and blue components in a sample of a test color image.

FIG. 4. Red in a further sample of the test color image
It is a figure showing each pixel value of a green and blue component.

FIG. 5 is a block diagram showing the pre-processing circuit shown in FIG. 2 in further detail;

FIG. 6 is a diagram for explaining the operation of a preprocessing circuit in the data compression encoder shown in FIG.

FIG. 7 is a diagram illustrating a DPCM modeling process.

FIG. 8 is a block diagram showing an internal configuration of the data compression decoder shown in FIG.

FIG. 9 is a block diagram showing a post-processing circuit of the data compression decoder shown in FIG. 8 in more detail;

FIG. 10 is a block diagram showing another specific example of the preprocessing circuit.

FIG. 11 is a flowchart illustrating an operation of the preprocessing circuit illustrated in FIG. 10;

FIG. 12 is a block diagram showing a configuration of a post-processing circuit corresponding to the pre-processing circuit shown in FIG.

Continuing on the front page (72) Inventor Perry Jason Charles UK Catty 130 Xda Brew Sally Waybridge Brooklands The Heights (no address) Sony United Kingdom Limited Limited F-term (reference) 5J064 AA02 BA04 BA09 BA11 BB03 BC08 BC27 BC28 BD03

Claims (39)

[Claims] 1. A data processing apparatus for expressing, as a modeled data symbol, a data symbol of a data source having data of a first component and data of a second component related to the data of the first component. By predicting each data symbol of the first component from the data symbol of the first component and the data symbol of the second component, a modeled data symbol representing the data symbol of the first component is generated. A data processing device comprising one prediction processing circuit. 2. The first prediction processing circuit calculates a difference between a symbol of a preceding first component and a symbol of a corresponding preceding second component, and calculates a difference between each of the symbols of the first component. The predicted value of each first component symbol is determined by subtracting the difference from the corresponding second component symbol, and corresponding to the first component symbol from the first component symbol. 2. The data processing apparatus according to claim 1, wherein a prediction error of each first component symbol is calculated by subtracting a prediction value, and the modeled data symbol is generated from the prediction error. 3. The data processing apparatus according to claim 2, wherein the modeled data symbol is generated from each of the prediction errors by using the alphabet size of the data symbol of the source data as a modulus. 4. The apparatus according to claim 1, further comprising a second prediction processing circuit configured to generate a second modeled data symbol representing the data symbol of the second component. Data processing device. 5. The second prediction processing circuit predicts each of the second component data symbols from a preceding second component data symbol, and calculates a difference between the original second data symbol and the prediction. 5. The data processing apparatus according to claim 4, wherein the second modeled data symbol is generated from the data. 6. The second prediction processing circuit, wherein the second prediction processing circuit is configured to calculate the second prediction symbol from at least one preceding second component symbol and at least one other second component symbol weighted by a corresponding weighting factor. Generating a prediction value of the symbol of the component; calculating a prediction error from a difference between the symbol of the second component and the prediction value of the symbol of each second component; By generating the second modeled data symbol from the prediction error,
6. The data processing device according to claim 5, wherein a second modeled data symbol representing the source data symbol is generated. 7. The data processing apparatus according to claim 4, wherein the second prediction processing circuit operates based on differential pulse code modulation. 8. A data processing for generating estimated values of source data symbols of first and second components from first modeled data symbols generated by the data processing apparatus according to claim 1. Description: In the apparatus, each first model symbol is subtracted from each second component symbol corresponding to each modeled data symbol by subtracting the difference between the preceding second data symbol and the preceding first data symbol. Generating a predicted value of each first component symbol corresponding to the converted data symbol; adding the predicted value of each first component data symbol to the corresponding modeled data symbol; A data processing device including an inverse processing circuit that generates an estimated value of a symbol. 9. The first device according to claim 4, wherein
And a first and second component source data symbol estimate from the second modeled data symbol and a preceding second component symbol estimate and a preceding at least one second symbol. Generating a predicted value of each second component symbol from a comparison with the estimated value of the second component data symbol;
Combining the second modeled data symbol with the predicted value of each of the second component symbols to generate an estimated value of each second component symbol to generate the second modeled data symbol from the second modeled data symbol. A first inverse processing circuit for generating an estimated value of the source data symbol of the second component, and a predicted value of each data symbol of the first component from the estimated value of the symbol of the first component; And a second inverse processing circuit for combining the predicted value of the modeled data symbol with the predicted value of the symbol of the first component. 10. The second inverse processing circuit according to claim 1, wherein a difference between a preceding second data symbol and a preceding first data symbol from each second component symbol corresponding to each first modeled data symbol. To generate a predicted value of the symbol of each first component corresponding to each first modeled data symbol, and predicting the symbol of each first component in the corresponding modeled data symbol 10. The data processing apparatus according to claim 9, wherein an estimated value of each of the first component symbols is generated by adding the values. 11. The second inverse prediction processing circuit modulo an alphabet size of a modeled data symbol.
11. The data processing apparatus according to claim 10, wherein the estimated value of the symbol of the first component is reproduced by adding the predicted value to the corresponding modeled data symbol. 12. A data processing method for processing a data symbol of a data source having data of a first component and data of a second component related to the data of the first component, comprising the steps of: Estimating a data symbol of each first component from the data symbol and the data symbol of the second component; and a difference between a predicted value of the symbol of each first component and a corresponding data symbol of the first component. Generating a first modeled data symbol representing a data symbol of the first component from the first component. 13. The data processing method according to claim 12, further comprising the step of generating a second modeled data symbol from the data symbol of the second component. 14. A method for predicting each second component data symbol from a preceding second component data symbol, wherein the second modeling is performed based on an error between the original second data symbol and a predicted value. 14. The data processing method according to claim 13, wherein the second modeled data symbol is generated from the data symbol of the second component by generating a data symbol. 15. A data compression encoding apparatus for generating compressed encoded data from a data source having data of a first component and data of a second component related to the data of the first component, wherein: A pre-processing circuit that generates a first modeled data symbol representing a symbol of the first component data from the component data of the first component; and a first modeled data symbol that is connected to the pre-processing circuit. A compression encoding processing circuit that generates the compressed encoded data by expressing the symbol of the second component as a compressed encoded data symbol, wherein the preprocessing circuit includes a data symbol of the preceding first component. And a predicted value of the data symbol of the first component is generated from the data symbol of the second component, and the symbol of the first component and the symbol of the first component are generated. A data compression encoding apparatus for generating each of the first modeled data symbols from a difference between predicted values corresponding to the first and second data symbols. 16. The pre-processing circuit calculates a difference between a preceding symbol of the first component and a symbol of a preceding second component corresponding to the symbol of the first component. The second corresponding to the symbol of component 1
By calculating the predicted value of each of the first component symbols by subtracting the difference from the data symbol of the component, the prediction corresponding to the data symbol of the first component is calculated from the data symbol of each first component. 16. The data compression encoding apparatus according to claim 15, wherein a prediction error is calculated by subtracting a value, and the first modeled data symbol is generated from the prediction error. 17. The data compression encoding apparatus according to claim 16, wherein said preprocessing circuit generates said modeled data symbol from each prediction error by using the alphabet size of said modeled data symbol as a modulus. . 18. The symbol of the second component from the symbol of the at least one preceding second component and the symbol of the at least one other second component weighted by a corresponding weighting factor. And a prediction error is obtained from a difference between the symbol of the second component and the predicted value of each of the symbols of the second component. From the prediction error of the symbol of each of the second components, The second modeled data symbol is generated by generating the second modeled data symbol.
18. The data compression encoding apparatus according to claim 15, wherein a second modeled data symbol representing a symbol of a component is generated. 19. Data compression for generating an estimated value of a first and second component source data symbol from a compressed and encoded data symbol generated by the data compression and encoding device according to any one of claims 15 to 17. A decoding device that receives the compressed and encoded data symbol and generates first and second modeled data symbols from the compressed and encoded data symbol; and a data compression and decoding process circuit, And a post-processing circuit for generating an estimated value of the first component symbol from the first modeled data symbol combined with the second component data symbol. 20. The post-processing circuit according to claim 1, wherein a difference between the estimated value of the data symbol of the preceding second component and the estimated value of the data symbol of the preceding first component is determined by a second symbol corresponding to each of the modeled data symbols. The predicted value is calculated by subtracting the predicted value of the first data symbol from the estimated value of the data symbol of the component, and the predicted value of the first data symbol is added to the corresponding modeled data symbol. 20. The data compression / decoding device according to claim 19, wherein an estimated value of the symbol is generated. 21. The post-processing circuit reproduces the estimated value of the first component symbol by adding the predicted value to the corresponding modeled data symbol modulo the alphabet size of the modeled data symbol. 21. The data compression / decoding device according to claim 20, wherein: 22. The data compression / decoding device according to claim 1,
8. An estimated value of the source data symbol of the first and second components is generated from the compressed and encoded data symbol generated by the data compression and encoding device according to item 8. Generating the first and second modeled data symbols from the symbols, wherein the post-processing circuit includes at least one preceding second component symbol and at least one other second weighted by a corresponding weighting factor. From the symbol of the component
And generating a data symbol of the second component by combining the predicted value of the data symbol of the second component with the second modeled data symbol. 19. The data compression / decoding device according to claim 18, wherein 23. A data compression encoding method for compressing and encoding data of a first component and data of a second component related to the data of the first component, comprising: Generating a first modeled data symbol representing a data symbol of one component; compressing and encoding the first modeled data symbol and the symbol of the second component to generate a compressed encoded data symbol; Generating the first modeled data symbols, wherein the difference between the preceding first component symbol and the preceding second component symbol is determined for each first component. Calculating the predicted value of each first component symbol by subtracting from the second component symbol corresponding to the symbol; and calculating the first component symbol from the first component symbol. Data compression encoding, comprising: calculating a prediction error of each first component symbol by subtracting a prediction value corresponding to the minute symbol; and generating the modeled data symbol from the prediction error. Method. 24. A data compression / decoding method for decoding a compression-encoded data symbol which has been compression-encoded by the data compression / encoding method according to claim 23, and calculating estimated values of first and second source data symbols. At a step of compression-encoding the first modeled data symbol and the second component symbol to generate a compressed encoded data symbol; and a symbol of the preceding second component and a symbol of the preceding first component. Calculating a predicted value for each modeled data symbol by subtracting the difference between the symbol from the second component symbol corresponding to the modeled data symbol; Generating an estimated value of each of the first component symbols by adding the predicted value of each of the first component symbols to the symbol; A data compression and decoding method comprising: 25. A data processing device for expressing a data symbol of a data source having data of mutually related first, second and third components as a modeled data symbol, said data processing device comprising: And generating a first modeled data symbol representing the data of the first component from the data of the third component, wherein the first modeled data symbol is derived from the second and third data symbols. A data processing device for representing an error between the predicted value of the first component symbol and the first data symbol. 26. The symbol of the preceding second component (G
＾) and the first relational metric generated from the difference between the preceding first component symbol (B ＾) and the preceding third component symbol (R ＾) and the preceding second component. A second relation metric generated from the difference between the first component symbol (G ＾) and the corresponding second component metric for each first component symbol. A third relation metric is determined from the difference between the symbol of the component (G) and the symbol of the preceding component (R 先行).
Is determined to be equal to the symbol of the component (G 成分), and if the preceding symbols of the third and second components are equal (R ＾ = G ＾), the symbol of the corresponding second component (G ＾) is determined. ) And the corresponding first relation metric (G ＾ -B
＾) to generate a predicted value of the symbol of the first component, and if the symbols of the preceding third and second components are not equal (R ＾ ≠ G ＾), the corresponding second Generating a predicted value of the symbol of the first component from a difference between the symbol of the first component and the ratio of the first and second relation metrics scaled by the third relation metric. The data processing device according to claim 25. 27. A data processing device according to claim 25 or 26, wherein said first data symbol and said second and third data symbols are connected to said data processing device and are converted into compression-encoded data symbols. A data compression encoding device comprising: a compression encoding circuit that performs compression encoding. 28. A data processing apparatus for generating estimated values of first, second, and third data symbols from compressed encoded data symbols generated by the data compression encoding apparatus according to claim 27. A compression decoding circuit for receiving the encoded data symbol and generating a first modeled data symbol and a symbol of the second and third components from the compression encoded data symbol; and A post-processing circuit for generating an estimate of each first component symbol from the first modeled data symbol combined with the second and third component symbols, the first component symbol Are estimated from the first modeled data symbol, the second and third component symbols, and the first component symbol. The data processing apparatus characterized by being produced by adding the prediction value corresponding to the component of the symbol. 29. The pre-processing circuit, wherein the first relational metric generated from a difference between a preceding second component symbol (G ＾) and a preceding first component symbol (B ＾). And a second relation metric generated from the difference between the preceding third component symbol (R ＾) and the preceding second component symbol (G ＾), and determining each first Determining, for the component symbols, a third relation metric from a difference between a corresponding third component symbol (R) and a corresponding second component symbol (G);
The symbol (R) of the preceding third component is preceded by the second
Is determined to be equal to the symbol of the component (G 成分), and if the preceding symbols of the third and second components are equal (R ＾ = G ＾), the symbol of the corresponding second component (G ＾) is determined. ) And the corresponding first relation metric (G ＾ -B
＾) to generate a predicted value of the symbol of the first component, and if the symbols of the preceding third and second components are not equal (R 、 G ＾), the corresponding second Generating a predicted value of the symbol of the first component from a difference between the symbol of the first component and the ratio of the first and second relation metrics scaled by the third relation metric. The data processing device according to claim 28. 30. The apparatus according to claim 25, wherein the data is a color image, and the first, second, and third components represent red, green, and blue components. Data processing equipment. 31. A data processing method for processing source data having interrelated first, second and third component data, comprising: a preceding second component symbol (G ＾) and a preceding first The first relation metric generated from the difference between the symbol of the component (B ＾) and the symbol of the preceding third component (R ＾) and the symbol of the preceding second component (G ＾)
Determining a second relational metric generated from the difference between: a symbol of the first component, a symbol of the corresponding third component (R), and a symbol of the corresponding second component. (G) determining a third relation metric from the difference between the second component metric and the preceding third component symbol (R ＾).
Determining whether the symbol of the third component is equal to the symbol of the component (G ＾), and if the symbol of the preceding third and second components is equal (R ＾ = G ＾), the symbol of the corresponding second component (G) and the corresponding first relation metric (G ＾ -B
＾) generating one of the modeled data symbols corresponding to the first component symbol from the difference between the first and second component symbols, if the preceding third and second component symbols are not equal (R ＾). {G}), predicting the first component symbol from the difference between the corresponding second component symbol and the ratio of the first and second relation metrics scaled by the third relation metric Generating a value; and generating the modeled data symbol from a difference between a predicted value of each first component data symbol and a corresponding original first component data symbol. Method. 32. A method according to claim 31, wherein the first and second data symbols are generated from the modeled data symbols generated by the data processing method.
And a data processing method for generating an estimate of the symbol of the third component, wherein, for each first modeled data symbol, a symbol of the preceding second component (G ＾) and a symbol of the preceding first component (G ＾). B ＾) and the first relational metric generated from the difference between
And a second relational metric generated from the difference between the preceding second component symbol (G ＾) and, for each first component symbol, a corresponding third component symbol Determining a third relation metric from the difference between the symbol (R) and the corresponding second component symbol (G);
Determining whether the symbol of the third component is equal to the symbol of the component (G ＾), and if the symbol of the preceding third and second components is equal (R ＾ = G ＾), the symbol of the corresponding second component (G) and the corresponding first relation metric (G ＾ -B
Generating the predicted value of the symbol of the first component from the difference between シ ン ボ ル) and the symbol of the preceding third and second components is not equal (R ＾ G ＾). Generating a predicted value of the first component symbol from a difference between a second component symbol and a ratio of the first and second relation metrics scaled by the third relation metric; Generating an estimate of each of the first component data symbols from a combination of the predicted value of the first component data symbol and the corresponding modeled data symbol. 33. A signal representing data generated by the data processing device or the data compression encoding device according to any one of claims 1, 12, 15, 19, 25, and 29. 34. A data recording device comprising a recording / reproducing medium on which a signal according to claim 29 is recorded. 35. A computer loaded into a computer and operating the computer as a data processing device, a data compression encoding device or a data compression decoding device according to any one of claims 1, 9, 15, 19, 25 and 29. Computer program having instructions executable by: 36. A computer program having computer-executable instructions loaded on a computer and causing the computer to perform the method of any one of claims 12, 19, 31, and 32. 37. A computer program product comprising a computer readable medium having recorded thereon an information signal representing the computer program according to claim 35. 38. A data compression encoder or data compression decoder described herein with reference to the accompanying drawings. 39. A data compression encoding method or data compression decoding method described herein with reference to the accompanying drawings.